Intel Bay Trail Graphics On Linux Are Slower Than Windows

[phoronix]

Phoronix: Intel Bay Trail Graphics On Linux Are Slower Than Windows

A few days ago my benchmarking revealed Windows 8.1 is outperforming Ubuntu Linux with the latest Intel open-source graphics drivers on Haswell hardware. I have since conducted tests on the Celeron N2820 NUC, and sadly, the better OpenGL performance is found with Microsoft's operating system.

http://www.phoronix.com/vr.php?view=20494

[dungeon]

Intel Bay Trail Graphics On Linux Are Slower Than Windows

I can say that is not always true, we can assume that those numbers told some story but that is not 100% true story (graphics drivers are not quilty) .

At least for OpenArena 0.8.5 and Xonotic 0.7, but i also tested UrbanTerror 0.4.1... and not on Intel but on Radeon 8400 so graphics does not matter... On Windows and Linux, but also with wine, with catalyst, fglrx and opensorce radeon driver... blah, blah...

So to approve this assumpation for example try running those three benchmarks in wine (preferably all on low and 32bit) and i bet they will have the same or better performance then linux native binaries .

That is my experience, those exe just better perform than elfs .

[dungeon]

Tested again now , so how much faster is exe running in wine vs native for me:

OpenArena 0.8.5 +5%
UrbanTerror 0.4.1 +16%
Xonotic the same

And to point out expected is also that wine behave slower when compared with Windows , so there is no doubt to me that those exe are faster .

[johnc]

The whole thing is kind of weird. I thought Intel was trying to be competitive in low-end places like Android and Chromebooks?

[GreatEmerald]

Tested again now , so how much faster is exe running in wine vs native for me:



And to point out expected is also that wine behave slower when compared with Windows , so there is no doubt to me that those exe are faster .

But how does that make any sense...

[dungeon]

But how does that make any sense...

That is a percentage difference i have, what is a reason of that difference exactly i don't know... i assume there is diference in compilers, compile options, maybe also some performance difference in ID Tech engine exists between the OSes, some feature in code is there for Windows but not for Linux, etc .

[dungeon]

When i doing Xonotic benchmark, first time fps is slower on Windows, then second and third time is faster. That does not happen for me on Linux, maybe some cache mechanisam is different i guess.

Also with UrbanTerror benchmark graphics are much brighter visibily on eye when i use exe both in Windows or under wine are the same, but linux native is not like that - not that bright, etc.

[supercilious]

The whole thing is kind of weird. I thought Intel was trying to be competitive in low-end places like Android and Chromebooks?

On Bay Trail devices running Android, Intel uses a proprietary driver, completely different from the normal Linux driver. Significantly higher performance than the Linux driver from what I can tell too, judging by the gaming performance of these devices. Also, the video acceleration supports both encoding and decoding (and decodes formats that stutter on Linux).

I have only tested two such devices, the Onda V975i and the Teclast x98, and both are running quad core Bay Trail CPUs at 2.2W (Turbo 4W) and they completely trounce the N2910 at 5W (Turbo 7.5W) in GPU performance. The Linux driver on the N2910 does not even manage H264 playback at 4K resolutions on a 1080p screen, which is something the Windows driver does fine. The Android devices have no problem with 4K at all, even at the highest bit-rates.

See the Q&A entry on Android-IA at 01.org:
"seems your distro includes binary drivers supporting es 3.1 and not based on Mesa, will future Intel Linux Graphics Driver Installer at 01.org be based on this closed drivers so supporting tesselation etc..

Until further notice, Android* on Intel(r) Architecture 4.4 and beyond will be supported by the closed source graphics driver."

[johnc]

On Bay Trail devices running Android, Intel uses a proprietary driver, completely different from the normal Linux driver. Significantly higher performance than the Linux driver from what I can tell too, judging by the gaming performance of these devices. Also, the video acceleration supports both encoding and decoding (and decodes formats that stutter on Linux).

Thanks for the information as that explains the discrepancy.

Then my next question becomes... why won't they provide a high-performing proprietary driver on Linux like AMD and nvidia do? Why should we accept less than optimal performance from the hardware we purchased?

[bridgman]

Cool...

"Linux users demand closed source driver"

(runs away)